CN104039004A - Method for heterogeneous user pilot frequency power optimal distribution in large-scale multi-input multi-output system - Google Patents

Abstract

The invention discloses a method for heterogeneous user pilot frequency power optimal distribution in a large-scale multi-input multi-output system. The method comprises a first step of generating a group of randomly distributed heterogeneous users, enabling each user to undergo independent channel information, and calculating a system downlink achievable speed; a second step of forming a Lagrangian function L by the system downlink achievable speed and total power constraint conditions; a third step of assuming lambda to be a Lagrangian multiplier, performing derivation of L on pilot frequency power rhok and lambda to obtain an expression of rhok related to the lambda and the channel information; a fourth step of obtaining the value range of lambda based on the expression obtained in the third step according to the restrictions that rhok is larger than 0 and smaller than the total power; a fifth step of performing binary search according to the value range of lambda obtained in the fourth step, and obtaining an optimal pilot frequency power distribution value. According to the method, different channel information of the heterogeneous user is utilized to achieve optimal distribution of pilot frequency power, and the overall data transmission performance of the system is improved.

Description

Isomery user pilot frequency power optimizing distribution method in extensive multi-input multi-output system

Technical field

The invention belongs to wireless communication technology field, specifically, relate to isomery user pilot frequency power optimizing distribution method in a kind of extensive multi-input multi-output system.

Background technology

Along with user constantly increases the demand of high-speed data service, and ever-increasing community user number, mobile communications network increases day by day to the demand of frequency spectrum resource, large-scale and multiple users multiple-input and multiple-output (multi-user's multiple-input and multiple-output, in literary composition, being called for short: MU-MIMO) system improves system spectrum utilance by increasing antenna for base station number, obtains extensive concern.Extensive MU-MIMO system is equipped with more than the more antenna of number of users in base station side, and community user is equipped with individual antenna, this a large amount of antenna for base station, in identical running time-frequency resource, serve multiple terminal uses simultaneously, utilize the pilot tone of uplink and the reciprocity of TDD (time division duplex) system channel, obtain all users' up-downgoing channel estimation value, thereby carry out downlink precoding.

The essential characteristic of extensive MU-MIMO system is that the number of antennas of base station side has increased by one more than magnitude than traditional MU-MIMO system.It,, than traditional MU-MIMO system, possesses distinctive advantage: the capacity that obtains higher multiple; Higher power utilization and the availability of frequency spectrum; Can use relatively inexpensive, lower powered device; Better link reliability etc.

Traditional extensive mimo system, in community, all users adopt orthogonal guide frequency, and base station utilizes these orthogonal guide frequencies and TDD system channel reciprocity to carry out channel estimating, thereby obtains all users' up-downgoing channel estimating information.But, due to the restriction of coherence time and number of users, identical orthogonal pilot frequency sequence need to be multiplexing in multiple communities, thereby base station will be subject to the interference of the pilot frequency information that co-frequency cell user sends in the time receiving ascending pilot frequency information, produces the problem of pilot pollution.For pilot pollution, mainly containing cooperates by minizone eliminates pilot pollution, alleviate the method such as pilot pollution and pilot pollution precoding by dispatch pilot tone on time and space at present.On the other hand, in extensive MU-MIMO system, user is generally isomery user, and each user experiences independently channel, thereby each user just may reach different descending achievable rates.So whether we consider to obtain by distributing to pilot power that isomery user is different the lifting of performance.The research of this respect is at present also fewer.

Summary of the invention

Technical problem to be solved by this invention is: isomery user pilot frequency power optimizing distribution method in a kind of extensive multi-input multi-output system is provided, the method is utilized the channel information that isomery user is different, in the situation that total pilot power is limited, the optimization that realizes each user's pilot power distributes, effectively utilize limited pilot power, promote thereby reach overall system data transmission performance.

For solving the problems of the technologies described above, the technical solution used in the present invention is:

Isomery user pilot frequency power optimizing distribution method in a kind of extensive multi-input multi-output system, the method comprises following process:

Step 1) there are a base station and K isomery user in community, use g _k=[g _1k, g _2k..., g _mk] ^trepresent in community that k user arrives the channel vector of base station, wherein, the antenna sum that M is base station, k=1,2,3 ... K, in community, the multiple transmission coefficient of the m root antenna of k user terminal to base station is g _mk, m is positive integer, and 1≤m≤M, h _mkrepresent the multiple rapid fading factor of the m root antenna of k user terminal to base station in community, β _kthe slow fading coefficient that represents k user terminal to base station in community, so-called isomery user, for different k, β _kinequality; User adopts method of estimation when long to obtain slow fading factor beta _k;

Step 2) first user carry out ascending pilot frequency transmission, uses ρ _krepresent the average pilot power that k user sends, in channel estimation phase, the reception signal of base station is:

Wherein k user's pilot signal, τ>=K, wherein each element of Z is obeyed adopt MMSE to estimate, can obtain:

Channel vector can be decomposed into channel estimation vector is the character of estimating according to MMSE, wherein ${\mathrm{\σ}}_k^2=\frac{{\mathrm{\τ\ρ}}_k{\mathrm{\β}}_k^2}{1+{\mathrm{\τ\ρ}}_k{\mathrm{\β}}_k},{\mathrm{\ϵ}}_k^2={\mathrm{\β}}_k-{\mathrm{\σ}}_k^2;$

Step 3) then base station carry out downlink data transmission.Suppose s _tbe to be transferred to t user's signal and Ε [| s _t| ²]=1.Base station utilizes channel estimating information to carry out linear predictive coding to the signal that will transmit, and supposes p _tbe t user's precoding vectors, the downstream signal of k user's reception is:

$y_k=\underset{t=1}{\overset{K}{\mathrm{\Σ}}}\sqrt{P_d}{g}_k^H{p}_t{s}_t+{\mathrm{\υ}}_k$ Formula (3)

Wherein P _dfor downlink data power, υ _kit is unit additive noise.As can be seen from the above equation, k user's down receiving signal is subject to the interference of other user's downlink datas;

Step 4) calculate k user's descending Signal to Interference plus Noise Ratio (SINR).Formula (3) is rewritten as:

Wherein, formula (4) has marked signal, interference and noise section.Thereby we can obtain k user's descending achievable rate:

${\mathrm{SINR}}_k=\frac{P_d{\left|g_k^H{p}_k\right|}^2}{\underset{t=1,t\≠k}{\overset{K}{\mathrm{\Σ}}}{P}_d{\left|g_k^H{p}_t\right|}^2+1}$ Formula (5)

Step 5) the descending achievable rate of computing system.Precoding vectors based on MF is:

$p_k=\frac{{\hat{g}}_k}{\sqrt{K}\left|\right|{\hat{g}}_k\left|\right|}=\frac{{\hat{g}}_k}{{\mathrm{\α}}_k\sqrt{\mathrm{MK}}}$ Formula (6)

Wherein, normalization factor, and count M increase with antenna for base station and remain unchanged, consideration downlink data power P in order to ensure the total transmitting power in base station _dwith mode along with M changes, wherein E _dconstant.Thereby the descending SINR of user k is tending towards infinite approximation at M and is:

$\begin{array}{c}\underset{M\→\∞}{\lim }{\mathrm{SINR}}_k=\underset{M\→\∞}{\lim }\frac{P_d{\left|g_k^H{p}_k\right|}^2}{\underset{t=1,t\≠k}{\overset{K}{\mathrm{\Σ}}}{P}_d{\left|g_k^H{p}_t\right|}^2+1}\\ =\underset{M\→\∞}{\lim }\frac{\frac{E_d}{{\mathrm{\α}}_k^{2}K}{\left|\frac{{({\hat{g}}_k+{\tilde{g}}_k)}^H{\hat{g}}_k}{M}\right|}^2}{\underset{t=1,t\≠k}{\overset{K}{\mathrm{\Σ}}}\frac{E_d}{{\mathrm{\α}}_t^{2}K}{\left|\frac{{({\hat{g}}_k+{\tilde{g}}_k)}^H{\hat{g}}_t}{M}\right|}^2+1}\\ =\frac{E_d}{K}\frac{{\mathrm{\τ\ρ}}_k{\mathrm{\β}}_k^2}{1+{\mathrm{\τ\ρ}}_k{\mathrm{\β}}_k}\end{array}$ Formula (7)

System descending achievable rate is:

$C^{\∞}\stackrel{\mathrm{\Δ}}{=}\underset{M\→\∞}{\lim }R=\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{\log }_2(1+\frac{E_d}{K}\frac{{\mathrm{\τ\ρ}}_k{\mathrm{\β}}_k^2}{1+{\mathrm{\τ\ρ}}_k{\mathrm{\β}}_k})$ Formula (8)

Step 6) utilize lagrange's method of multipliers Solve problems.Suppose that P is total pilot power, first problem represented with the following formula of optimizing:

$\begin{array}{cc}\underset{{\mathrm{\ρ}}_k,k=\mathrm{1,2},\·\·\·,K}{\min }& -C^{\∞}\\ s.t.& \underset{k=1}{\overset{K}{\mathrm{\Σ}}}{\mathrm{\ρ}}_k\≤P\end{array}$ Formula (9)

Wherein, without loss of generality, by the τ ρ in formula (8) _kuse ρ _krepresent.Suppose that λ is non-negative Lagrange factor, structure Lagrangian is as follows:

$L=-\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{\log }_2(1+\frac{E_d}{K}\frac{{\mathrm{\ρ}}_k{\mathrm{\β}}_k^2}{1+{\mathrm{\ρ}}_k{\mathrm{\β}}_k})+\mathrm{\λ}(\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{\mathrm{\ρ}}_k-P)$ Formula (10)

By L about ρ _kcan obtain with λ differentiate:

$\left\{\begin{array}{c}\frac{\∂L}{{\∂\mathrm{\ρ}}_k}=\mathrm{\λ}-\frac{E_d{\mathrm{\β}}_k^2}{\ln 2(K+K{\mathrm{\ρ}}_k{\mathrm{\β}}_k+{\mathrm{\ρ}}_k{\mathrm{\β}}_k^2{E}_d)(1+{\mathrm{\ρ}}_k{\mathrm{\β}}_k)}\\ \frac{\∂L}{\∂\mathrm{\λ}}=\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{\mathrm{\ρ}}_k-P\end{array}\right.$ Formula (11)

Make formula (11) thus be that 0 expression formula that obtains each user's pilot power is:

${\mathrm{\ρ}}_k=\frac{\sqrt{E_d^2+\frac{{4E}_{d}K}{\mathrm{\λ}\ln 2}+\frac{4E_d^2{\mathrm{\β}}_k}{\mathrm{\λ}\ln 2}}}{2(K+E_d{\mathrm{\β}}_k)}-\frac{(2K+E_d{\mathrm{\β}}_k)}{2({\mathrm{\β}}_{k}K+E_d{\mathrm{\β}}_k^2)}$ Formula (12)

Step 7) utilize binary search to ask the optimal value of isomery user pilot power.Utilize 0≤ρ _kthe Power Limitation condition of≤P, tries to achieve lower limit and the upper limit { λ of one group of λ based on formula (12) _{k, max}and { λ _{k, min}, the desired value λ of λ ^*be positioned at following interval: the pilot power that ensures each user is no more than P.Make λ _land λ _ube respectively lower limit and the upper limit of the each iteration of binary search, λ ₀be the λ value that each iteration adopts, the process of binary search is as follows:

I. initialization ${\mathrm{\λ}}_l=\underset{k}{\max }\left\{{\mathrm{\λ}}_{k,\min }\right\}$ With ${\mathrm{\λ}}_u=\underset{k}{\max }\left\{{\mathrm{\λ}}_{k,\max }\right\}.$

Ii. order solve ρ according to formula (12) _k, and calculate total pilot power

If iii. P _all> P, makes λ _l=λ ₀, otherwise make λ _u=λ ₀.

Iv. jump to ii. until meet P-P _all< ∈.Wherein ∈ is the limits of error.

In said process, if the pilot power value calculating for user k is less than zero, be set to zero.

Further, step 7) in adopt binary search can reduce in the following way mean iterative number of time:

A. by { λ _{k, max}arrange from small to large order ${\mathrm{\&lambda;}}_l=\underset{k}{\max }\left\{{\mathrm{\&lambda;}}_{k,\min }\right\}$ With ${\mathrm{\&lambda;}}_u=\underset{k}{\min }\left\{{\mathrm{\&lambda;}}_{k,\max }\right\}.$

B. get successively { λ _{k, max}in one value calculate ρ _k, and calculate total pilot power

If c. P _all< P, carries out as claim 1 step 7 in current scope) in binary search process, otherwise make λ _l=λ _u, λ _ufor from { λ _{k, max}extract value, jump to b.

In addition, step 1) to step 7) in pilot power allocation method be all the scene for single community, for many cell scenario, can adopt similar method to carry out isomery user's pilot frequency power optimizing.According to similar step 1) to step 5) the formula of the descending achievable rate of process computation:

$\underset{M\&RightArrow;\&infin;}{\lim }{\mathrm{SINR}}_{\mathrm{ik}}=\frac{\frac{E_d}{K}\left(\frac{{\mathrm{\&tau;\&rho;}}_k{\left({\mathrm{\&beta;}}_{\mathrm{ik}}^i\right)}^2}{1+\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_{\mathrm{lk}}^i}\right)}{\underset{j=1,j\&NotEqual;i}{\overset{L}{\mathrm{\&Sigma;}}}\frac{E_d}{K}\left(\frac{{\mathrm{\&tau;\&rho;}}_k{\left({\mathrm{\&beta;}}_{\mathrm{ik}}^j\right)}^2}{1+\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_{\mathrm{lk}}^j}\right)+1}$ Formula (13)

Wherein, represent in l community the slow fading coefficient of k user terminal to i base station.Here think that the user of simultaneously transmitting pilot tone in all communities has identical pilot power.Formula (13) is simplified, is obtained:

$\underset{M\&RightArrow;\&infin;}{\lim }{\mathrm{SINR}}_{\mathrm{ik}}=\frac{\frac{E_d}{K}{\mathrm{\&rho;}}_k{\left({\mathrm{\&beta;}}_{\mathrm{ik}}^i\right)}^2}{{\mathrm{\&rho;}}_k(\underset{j=1,j\&NotEqual;i}{\overset{L}{\mathrm{\&Sigma;}}}\frac{E_d}{K}{\left({\mathrm{\&beta;}}_{\mathrm{ik}}^j\right)}^2+\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}_{\mathrm{lk}}^i)+1}$ Formula (14)

Then according to step 6) and step 7) mode construct lagrangian optimization problem, and obtain the optimal value of pilot power by binary search.

Compared with prior art, the present invention has following beneficial effect:

(1) take full advantage of the different channel information of isomery user in extensive MU-MIMO system, the optimization that realizes each user's pilot power distributes, and promotes thereby reach overall system data transmission performance, has certain practicality.

(2) scheme that the present invention proposes, in the situation that total pilot power is limited, effectively utilizes limited pilot power, promotes thereby reach overall system data transmission performance.

Brief description of the drawings

Fig. 1 is that the descending achievable rate of pilot power allocation method of the present invention and traditional mean allocation method is about the comparative graph of total pilot power.

Fig. 2 be pilot power allocation method of the present invention with traditional mean allocation method under two groups of different user distributions of channel vector variance descending achievable rate about the comparative graph of total pilot power.

Fig. 3 is each isomery user achievable rate schematic diagram in pilot power allocation method of the present invention and traditional mean allocation method.

Fig. 4 is that the descending achievable rate of pilot power allocation method of the present invention and traditional mean allocation method under two cell scenario is about the comparative graph of total pilot power.

Fig. 5 (a) is each isomery user achievable rate schematic diagram in first community pilot power allocation method of the present invention and traditional mean allocation method under two cell scenario.

Fig. 5 (b) is each isomery user achievable rate schematic diagram in two cell scenario lower second community pilot power allocation method of the present invention and traditional mean allocation method.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention is further elaborated.

Isomery user pilot frequency power optimizing distribution method in the extensive multi-input multi-output system of one of the present invention, this scheme comprises following process:

Step 1) there are a base station and K isomery user in community, use g _k=[g _1k, g _2k..., g _mk] ^trepresent in community that k user arrives the channel vector of base station, wherein, the antenna sum that M is base station, k=1,2,3 ... K, in community, the multiple transmission coefficient of the m root antenna of k user terminal to base station is g _mk, m is positive integer, and 1≤m≤M, h _mkrepresent the multiple rapid fading factor of the m root antenna of k user terminal to base station in community, β _kthe slow fading coefficient that represents k user terminal to base station in community, so-called isomery user, for different k, β _kinequality; User adopts method of estimation when long to obtain slow fading factor beta _k;

Step 2) first user carry out ascending pilot frequency transmission, uses ρ _krepresent the average pilot power that k user sends, in channel estimation phase, the reception signal of base station is:

Wherein k user's pilot signal, τ>=K, wherein each element of Z is obeyed adopt MMSE to estimate, can obtain:

Channel vector can be decomposed into channel estimation vector is the character of estimating according to MMSE, wherein ${\mathrm{\&sigma;}}_k^2=\frac{{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_k^2}{1+{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_k},{\mathrm{\&epsiv;}}_k^2={\mathrm{\&beta;}}_k-{\mathrm{\&sigma;}}_k^2;$

Step 3) then base station carry out downlink data transmission.Suppose s _tbe to be transferred to t user's signal and Ε [| s _t| ²]=1.Base station utilizes channel estimating information to carry out linear predictive coding to the signal that will transmit, and supposes p _tbe t user's precoding vectors, the downstream signal of k user's reception is:

$y_k=\underset{t=1}{\overset{K}{\mathrm{\&Sigma;}}}\sqrt{P_d}{g}_k^H{p}_t{s}_t+{\mathrm{\&upsi;}}_k$ Formula (3)

Wherein P _dfor downlink data power, υ _kit is unit additive noise.As can be seen from the above equation, k user's down receiving signal is subject to the interference of other user's downlink datas;

Step 4) calculate k user's descending Signal to Interference plus Noise Ratio (SINR).Formula (3) is rewritten as:

Wherein, formula (4) has marked signal, interference and noise section.Thereby we can obtain k user's descending achievable rate:

${\mathrm{SINR}}_k=\frac{P_d{\left|g_k^H{p}_k\right|}^2}{\underset{t=1,t\&NotEqual;k}{\overset{K}{\mathrm{\&Sigma;}}}{P}_d{\left|g_k^H{p}_t\right|}^2+1}$ Formula (5)

Step 5) the descending achievable rate of computing system.Precoding vectors based on MF is:

$p_k=\frac{{\hat{g}}_k}{\sqrt{K}\left|\right|{\hat{g}}_k\left|\right|}=\frac{{\hat{g}}_k}{{\mathrm{\&alpha;}}_k\sqrt{\mathrm{MK}}}$ Formula (6)

Wherein, normalization factor, and count M increase with antenna for base station and remain unchanged, consideration downlink data power P in order to ensure the total transmitting power in base station _dwith mode along with M changes, wherein E _dconstant.Thereby the descending SINR of user k is tending towards infinite approximation at M and is:

$\begin{array}{c}\underset{M\&RightArrow;\&infin;}{\lim }{\mathrm{SINR}}_k=\underset{M\&RightArrow;\&infin;}{\lim }\frac{P_d{\left|g_k^H{p}_k\right|}^2}{\underset{t=1,t\&NotEqual;k}{\overset{K}{\mathrm{\&Sigma;}}}{P}_d{\left|g_k^H{p}_t\right|}^2+1}\\ =\underset{M\&RightArrow;\&infin;}{\lim }\frac{\frac{E_d}{{\mathrm{\&alpha;}}_k^{2}K}{\left|\frac{{({\hat{g}}_k+{\tilde{g}}_k)}^H{\hat{g}}_k}{M}\right|}^2}{\underset{t=1,t\&NotEqual;k}{\overset{K}{\mathrm{\&Sigma;}}}\frac{E_d}{{\mathrm{\&alpha;}}_t^{2}K}{\left|\frac{{({\hat{g}}_k+{\tilde{g}}_k)}^H{\hat{g}}_t}{M}\right|}^2+1}\\ =\frac{E_d}{K}\frac{{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_k^2}{1+{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_k}\end{array}$ Formula (7)

System descending achievable rate is:

$C^{\&infin;}\stackrel{\mathrm{\&Delta;}}{=}\underset{M\&RightArrow;\&infin;}{\lim }R=\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\log }_2(1+\frac{E_d}{K}\frac{{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_k^2}{1+{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_k})$ Formula (8)

Step 6) utilize lagrange's method of multipliers Solve problems.Suppose that P is total pilot power, first problem represented with the following formula of optimizing:

$\begin{array}{cc}\underset{{\mathrm{\&rho;}}_k,k=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;,K}{\min }& -C^{\&infin;}\\ s.t.& \underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_k\&le;P\end{array}$ Formula (9)

Wherein, without loss of generality, by the τ ρ in formula (8) _kuse ρ _krepresent.Suppose that λ is non-negative Lagrange factor, structure Lagrangian is as follows:

$L=-\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\log }_2(1+\frac{E_d}{K}\frac{{\mathrm{\&rho;}}_k{\mathrm{\&beta;}}_k^2}{1+{\mathrm{\&rho;}}_k{\mathrm{\&beta;}}_k})+\mathrm{\&lambda;}(\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_k-P)$ Formula (10)

By L about ρ _kcan obtain with λ differentiate:

$\left\{\begin{array}{c}\frac{\&PartialD;L}{{\&PartialD;\mathrm{\&rho;}}_k}=\mathrm{\&lambda;}-\frac{E_d{\mathrm{\&beta;}}_k^2}{\ln 2(K+K{\mathrm{\&rho;}}_k{\mathrm{\&beta;}}_k+{\mathrm{\&rho;}}_k{\mathrm{\&beta;}}_k^2{E}_d)(1+{\mathrm{\&rho;}}_k{\mathrm{\&beta;}}_k)}\\ \frac{\&PartialD;L}{\&PartialD;\mathrm{\&lambda;}}=\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_k-P\end{array}\right.$ Formula (11)

Make formula (11) thus be that 0 expression formula that obtains each user's pilot power is:

${\mathrm{\&rho;}}_k=\frac{\sqrt{E_d^2+\frac{{4E}_{d}K}{\mathrm{\&lambda;}\ln 2}+\frac{4E_d^2{\mathrm{\&beta;}}_k}{\mathrm{\&lambda;}\ln 2}}}{2(K+E_d{\mathrm{\&beta;}}_k)}-\frac{(2K+E_d{\mathrm{\&beta;}}_k)}{2({\mathrm{\&beta;}}_{k}K+E_d{\mathrm{\&beta;}}_k^2)}$ Formula (12)

Step 7) utilize binary search to ask the optimal value of isomery user pilot power.Utilize 0≤ρ _kthe Power Limitation condition of≤P, tries to achieve lower limit and the upper limit { λ of one group of λ based on formula (12) _{k, max}and { λ _{k, min}, the desired value λ of λ ^*be positioned at following interval: the pilot power that ensures each user is no more than P.Make λ _land λ _ube respectively lower limit and the upper limit of the each iteration of binary search, λ ₀be the λ value that each iteration adopts, the process of binary search is as follows:

I. initialization ${\mathrm{\&lambda;}}_l=\underset{k}{\max }\left\{{\mathrm{\&lambda;}}_{k,\min }\right\}$ With ${\mathrm{\&lambda;}}_u=\underset{k}{\max }\left\{{\mathrm{\&lambda;}}_{k,\max }\right\}.$

Ii. order solve ρ according to formula (12) _k, and calculate total pilot power

If iii. P _all> P, makes λ _l=λ ₀, otherwise make λ _u=λ ₀.

Iv. jump to ii. until meet P-P _all< ∈.Wherein ∈ is the limits of error.

In said process, if the pilot power value calculating for user k is less than zero, be set to zero.

Further, step 7) in adopt binary search can reduce in the following way mean iterative number of time:

A. by { λ _{k, max}arrange from small to large order ${\mathrm{\&lambda;}}_l=\underset{k}{\max }\left\{{\mathrm{\&lambda;}}_{k,\min }\right\}$ With ${\mathrm{\&lambda;}}_u=\underset{k}{\min }\left\{{\mathrm{\&lambda;}}_{k,\max }\right\}.$

B. get successively { λ _{k, max}in one value calculate ρ _k, and calculate total pilot power

If c. P _all< P, carries out as claim 1 step 7 in current scope) in binary search process, otherwise make λ _l=λ _u, λ _ufor from { λ _{k, max}extract value, jump to b.

In addition, step 1) to step 7) in pilot power allocation method be all the scene for single community, for many cell scenario, can adopt similar method to carry out isomery user's pilot frequency power optimizing.According to step 1 in similar 1) to step 5) the formula of the descending achievable rate of process computation:

$\underset{M\&RightArrow;\&infin;}{\lim }{\mathrm{SINR}}_{\mathrm{ik}}=\frac{\frac{E_d}{K}\left(\frac{{\mathrm{\&tau;\&rho;}}_k{\left({\mathrm{\&beta;}}_{\mathrm{ik}}^i\right)}^2}{1+\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_{\mathrm{lk}}^i}\right)}{\underset{j=1,j\&NotEqual;i}{\overset{L}{\mathrm{\&Sigma;}}}\frac{E_d}{K}\left(\frac{{\mathrm{\&tau;\&rho;}}_k{\left({\mathrm{\&beta;}}_{\mathrm{ik}}^j\right)}^2}{1+\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_{\mathrm{lk}}^j}\right)+1}$ Formula (13)

Wherein, represent in l community the slow fading coefficient of k user terminal to i base station.Here think that the user of simultaneously transmitting pilot tone in all communities has identical pilot power.Formula (13) is simplified, is obtained:

$\underset{M\&RightArrow;\&infin;}{\lim }{\mathrm{SINR}}_{\mathrm{ik}}=\frac{\frac{E_d}{K}{\mathrm{\&rho;}}_k{\left({\mathrm{\&beta;}}_{\mathrm{ik}}^i\right)}^2}{{\mathrm{\&rho;}}_k(\underset{j=1,j\&NotEqual;i}{\overset{L}{\mathrm{\&Sigma;}}}\frac{E_d}{K}{\left({\mathrm{\&beta;}}_{\mathrm{ik}}^j\right)}^2+\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}_{\mathrm{lk}}^i)+1}$ Formula (14)

Then according to step 6) and step 7) mode construct lagrangian optimization problem, and obtain the optimal value of pilot power by binary search.

The above is only the preferred embodiment of the present invention; be noted that for those skilled in the art; under the premise without departing from the principles of the invention, can also make some improvements and modifications, these improvements and modifications also should be considered as protection scope of the present invention.

Of the present inventionly take full advantage of the different channel information of isomery user in extensive MU-MIMO system, in the situation that total pilot power is limited, the optimization that realizes each user's pilot power distributes, effectively utilize limited pilot power, promote thereby reach overall system data transmission performance, there is certain practicality.

Enumerate some examples below, illustrate the premium properties that technical scheme of the present invention has.

L-G simulation test 1

Simulating scenes parameter: establishing radius of society is 500m, and base station is positioned at center of housing estate, user is evenly distributed on apart from base station at least in the cell range of 35cm, number of users K=10; Large scale fading factor model is that on average decline how much declines and the standard deviation of exponent gamma=3.8dB is σ _shadowthe shadow fading of the logarithm normal distribution of=8dB.In test, adopt monte carlo method, produce at random 5000 times independently user distribution carry out emulation, simulation result is 5000 times average.

Fig. 1 has provided downlink data power and has been respectively the descending achievable rate that adopts pilot power allocation method of the present invention and traditional mean allocation method under 10dB and 20dB condition about two groups of comparative graph of total pilot power.In Fig. 1, abscissa represents total pilot power, the dB of unit, and ordinate represents descending achievable rate, the bps/Hz of unit.In figure, adding asterisk solid line is the achievable rate curve of pilot distribution method of the present invention, and dotted line is the achievable rate curve of traditional mean allocation method.As can be seen from Figure 1, under this simulating scenes, total pilot power is from 0dB to 20dB, and the descending achievable rate that pilot power allocation method of the present invention obtains is all high than traditional mean allocation method.Along with the increase of total pilot power, performance gain diminishes.Meanwhile, large when pilot power allocation method of the present invention obtains when downlink data power 20dB descending achievable rate ratio of gains downlink data power 10dB.This shows under this scene, and pilot power allocation method of the present invention is compared mean allocation method and reached performance boost.

L-G simulation test 2

Simulating scenes parameter: establishing radius of society is 500m, and base station is positioned at center of housing estate, user is evenly distributed on apart from base station at least in the cell range of 35cm, number of users K=10; Large scale fading factor model is that on average decline how much declines and the standard deviation of exponent gamma=3.8dB is σ _shadowthe shadow fading of the logarithm normal distribution of=8dB.In test, produce at random two groups of users that channel vector variance is different, two groups of user { β of emulation here _kvariance be respectively 0.0919 and 0.1543.

Fig. 2 provided pilot power allocation method of the present invention and traditional mean allocation method under two groups of different user distributions of channel vector variance descending achievable rate about the comparative graph of total pilot power.Abscissa represents total pilot power, the dB of unit, and ordinate represents descending achievable rate, the bps/Hz of unit.What dotted line represented is, and one group of user that channel vector variance is larger adopts the up achievable rate curve of pilot distribution method of the present invention and average mark method of completing the square, adds one group of user that asterisk solid line represents that road vector variance is less and adopt the up achievable rate curve of pilot distribution method of the present invention and average mark method of completing the square.As can be seen from Figure 2 the up achievable rate gain that one group of user that, channel vector variance is larger adopts pilot distribution method of the present invention to obtain is more obvious.

L-G simulation test 3

Simulating scenes parameter: establishing radius of society is 500m, and base station is positioned at center of housing estate, user is evenly distributed on apart from base station at least in the cell range of 35cm, number of users K=5; Large scale fading factor model is that on average decline how much declines and the standard deviation of exponent gamma=3.8dB is σ _shadowthe shadow fading of the logarithm normal distribution of=8dB.In test, adopt monte carlo method, produce at random 5000 times independently user distribution carry out emulation, simulation result is 5000 times average.Downlink data power is 20dB, and total pilot power is 10dB.

Fig. 3 has provided each isomery user achievable rate schematic diagram in pilot power allocation method of the present invention and traditional mean allocation method.In Fig. 3, abscissa represents user index, and ordinate represents descending achievable rate.In figure, the large scale fading factor that 5 users' index is corresponding with it is inversely proportional to.As can be seen from Figure 3, the lifting that pilot power allocation method of the present invention is obtained system descending achievable rate by sacrificing the user's that channel fading is larger pilot power.

L-G simulation test 4

Simulating scenes parameter: be provided with Liang Ge community, radius of society is 500m, and base station is positioned at center of housing estate, user is evenly distributed on apart from base station at least in the cell range of 35cm, number of users K=10; Large scale fading factor model is that on average decline how much declines and the standard deviation of exponent gamma=3.8dB is σ _shadowthe shadow fading of the logarithm normal distribution of=8dB.In test, adopt monte carlo method, produce at random 5000 times independently user distribution carry out emulation, simulation result is 5000 times average.

Fig. 4 has provided downlink data power under two cell scenario and has been respectively the descending achievable rate that adopts pilot power allocation method of the present invention and traditional mean allocation method under 10dB and 20dB condition about two groups of comparative graph of total pilot power.In Fig. 4, abscissa represents total pilot power, the dB of unit, and ordinate represents descending achievable rate, the bps/Hz of unit.In figure, adding asterisk solid line is the achievable rate curve of pilot distribution method of the present invention, and dotted line is the achievable rate curve of traditional mean allocation method.As can be seen from Figure 4, under this simulating scenes, total pilot power is from 0dB to 20dB, and the descending achievable rate that pilot power allocation method of the present invention obtains is all high than traditional mean allocation method.Along with the increase of total pilot power, performance gain diminishes.Meanwhile, large when pilot power allocation method of the present invention obtains when downlink data power 20dB descending achievable rate ratio of gains downlink data power 10dB.This shows under this scene, and pilot power allocation method of the present invention is compared mean allocation method and reached performance boost.

L-G simulation test 5

Simulating scenes parameter: be provided with Liang Ge community, radius of society is 500m, and base station is positioned at center of housing estate, user is evenly distributed on apart from base station at least in the cell range of 35cm, number of users K=5; Large scale fading factor model is that on average decline how much declines and the standard deviation of exponent gamma=3.8dB is σ _shadowthe shadow fading of the logarithm normal distribution of=8dB.In test, adopt monte carlo method, produce at random 5000 times independently user distribution carry out emulation, simulation result is 5000 times average.Downlink data power is 20dB, and total pilot power is 10dB.

Fig. 5 has provided under two cell scenario each isomery user achievable rate schematic diagram in pilot power allocation method of the present invention and traditional mean allocation method, the schematic diagram that Fig. 5 (a) is first community, Fig. 5 (b) is the schematic diagram of second community.In Fig. 5, abscissa represents user index, and ordinate represents descending achievable rate.In figure, the large scale fading factor that 5 users' index is corresponding with it is inversely proportional to.As can be seen from Figure 5, the lifting that pilot power allocation method of the present invention is obtained system descending achievable rate by sacrificing the user's that channel fading is larger pilot power.

Taking above-mentioned foundation desirable embodiment of the present invention as enlightenment, by above-mentioned description, relevant staff can, not departing from the scope of this invention technological thought, carry out various change and amendment completely.The technical scope of this invention is not limited to the content on specification, must determine its technical scope according to claim scope.

Claims (4)

1. isomery user's a pilot frequency power optimizing distribution method in extensive multi-input multi-output system, is characterized in that, this distribution method comprises the following steps:

Step 1) there are a base station and K isomery user in community, use g _k=[g _1k, g _2k..., g _mk] ^trepresent in community that k user arrives the channel vector of base station, wherein, the antenna sum that M is base station, k=1,2,3 ... K, in community, the multiple transmission coefficient of the m root antenna of k user terminal to base station is g _mk, m is positive integer, and 1≤m≤M, h _mkrepresent the multiple rapid fading factor of the m root antenna of k user terminal to base station in community, β _kthe slow fading coefficient that represents k user terminal to base station in community, so-called isomery user, for different k, β _kinequality; User adopts method of estimation when long to obtain slow fading factor beta _k;

Step 2) first user carry out ascending pilot frequency transmission, uses ρ _krepresent the average pilot power that k user sends, in channel estimation phase, the reception signal of base station is:

Wherein k user's pilot signal, τ>=K, wherein each element of Z is obeyed adopt MMSE to estimate, can obtain:

Channel vector can be decomposed into channel estimation vector is the character of estimating according to MMSE, wherein ${\mathrm{\&sigma;}}_k^2=\frac{{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_k^2}{1+{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_k},{\mathrm{\&epsiv;}}_k^2={\mathrm{\&beta;}}_k-{\mathrm{\&sigma;}}_k^2;$

Step 3) then base station carry out downlink data transmission; Suppose s _tbe to be transferred to t user's signal and Ε [| s _t| ²]=1, base station utilizes channel estimating information to carry out linear predictive coding to the signal that will transmit, and supposes p _tbe t user's precoding vectors, the downstream signal of k user's reception is:

$y_k=\underset{t=1}{\overset{K}{\mathrm{\&Sigma;}}}\sqrt{P_d}{g}_k^H{p}_t{s}_t+{\mathrm{\&upsi;}}_k$ Formula (3)

Wherein P _dfor downlink data power, υ _kbe unit additive noise, can be found out by above formula (3), k user's down receiving signal is subject to the interference of other user's downlink datas;

Step 4) calculate k user's descending Signal to Interference plus Noise Ratio (SINR), formula (3) is rewritten as:

Wherein, formula (4) has marked signal, interference and noise section, thereby the descending achievable rate that obtains k user is:

${\mathrm{SINR}}_k=\frac{P_d{\left|g_k^H{p}_k\right|}^2}{\underset{t=1,t\&NotEqual;k}{\overset{K}{\mathrm{\&Sigma;}}}{P}_d{\left|g_k^H{p}_t\right|}^2+1}$ Formula (5)

Step 5) the descending achievable rate of computing system; Precoding vectors based on MF is:

$p_k=\frac{{\hat{g}}_k}{\sqrt{K}\left|\right|{\hat{g}}_k\left|\right|}=\frac{{\hat{g}}_k}{{\mathrm{\&alpha;}}_k\sqrt{\mathrm{MK}}}$ Formula (6)

Wherein, normalization factor, and count M increase with antenna for base station and remain unchanged, consideration downlink data power P in order to ensure the total transmitting power in base station _dwith mode along with M changes, wherein E _dconstant, thus the descending SINR of user k is tending towards infinite approximation at M and is:

$\begin{array}{c}\underset{M\&RightArrow;\&infin;}{\lim }{\mathrm{SINR}}_k=\underset{M\&RightArrow;\&infin;}{\lim }\frac{P_d{\left|g_k^H{p}_k\right|}^2}{\underset{t=1,t\&NotEqual;k}{\overset{K}{\mathrm{\&Sigma;}}}{P}_d{\left|g_k^H{p}_t\right|}^2+1}\\ =\underset{M\&RightArrow;\&infin;}{\lim }\frac{\frac{E_d}{{\mathrm{\&alpha;}}_k^{2}K}{\left|\frac{{({\hat{g}}_k+{\tilde{g}}_k)}^H{\hat{g}}_k}{M}\right|}^2}{\underset{t=1,t\&NotEqual;k}{\overset{K}{\mathrm{\&Sigma;}}}\frac{E_d}{{\mathrm{\&alpha;}}_t^{2}K}{\left|\frac{{({\hat{g}}_k+{\tilde{g}}_k)}^H{\hat{g}}_t}{M}\right|}^2+1}\\ =\frac{E_d}{K}\frac{{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_k^2}{1+{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_k}\end{array}$ Formula (7)

System descending achievable rate is:

$C^{\&infin;}\stackrel{\mathrm{\&Delta;}}{=}\underset{M\&RightArrow;\&infin;}{\lim }R=\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\log }_2(1+\frac{E_d}{K}\frac{{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_k^2}{1+{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_k})$ Formula (8)

Step 6) utilize lagrange's method of multipliers Solve problems; Suppose that P is total pilot power, first problem represented with the following formula of optimizing:

$\begin{array}{cc}\underset{{\mathrm{\&rho;}}_k,k=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;,K}{\min }& -C^{\&infin;}\\ s.t.& \underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_k\&le;P\end{array}$ Formula (9)

Wherein, by the τ ρ in formula (8) _kuse ρ _krepresent, suppose that λ is non-negative Lagrange factor, structure Lagrangian is as follows:

$L=-\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\log }_2(1+\frac{E_d}{K}\frac{{\mathrm{\&rho;}}_k{\mathrm{\&beta;}}_k^2}{1+{\mathrm{\&rho;}}_k{\mathrm{\&beta;}}_k})+\mathrm{\&lambda;}(\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_k-P)$ Formula (10)

By L about ρ _kcan obtain with λ differentiate:

$\left\{\begin{array}{c}\frac{\&PartialD;L}{{\&PartialD;\mathrm{\&rho;}}_k}=\mathrm{\&lambda;}-\frac{E_d{\mathrm{\&beta;}}_k^2}{\ln 2(K+K{\mathrm{\&rho;}}_k{\mathrm{\&beta;}}_k+{\mathrm{\&rho;}}_k{\mathrm{\&beta;}}_k^2{E}_d)(1+{\mathrm{\&rho;}}_k{\mathrm{\&beta;}}_k)}\\ \frac{\&PartialD;L}{\&PartialD;\mathrm{\&lambda;}}=\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_k-P\end{array}\right.$ Formula (11)

Make formula (11) thus be that 0 expression formula that obtains each user's pilot power is:

${\mathrm{\&rho;}}_k=\frac{\sqrt{E_d^2+\frac{{4E}_{d}K}{\mathrm{\&lambda;}\ln 2}+\frac{4E_d^2{\mathrm{\&beta;}}_k}{\mathrm{\&lambda;}\ln 2}}}{2(K+E_d{\mathrm{\&beta;}}_k)}-\frac{(2K+E_d{\mathrm{\&beta;}}_k)}{2({\mathrm{\&beta;}}_{k}K+E_d{\mathrm{\&beta;}}_k^2)}$ Formula (12)

Step 7) utilize binary search to ask the optimal value of isomery user pilot power; Utilize 0≤ρ _kthe Power Limitation condition of≤P, tries to achieve lower limit and the upper limit { λ of one group of λ based on formula (12) _{k, max}and { λ _{k, min}, the desired value λ of λ ^*be positioned at following interval: the pilot power that ensures each user is no more than P, makes λ _land λ _ube respectively lower limit and the upper limit of the each iteration of binary search, λ ₀be the λ value that each iteration adopts, the process of binary search is as follows:

I. initialization ${\mathrm{\&lambda;}}_l=\underset{k}{\max }\left\{{\mathrm{\&lambda;}}_{k,\min }\right\}$ With ${\mathrm{\&lambda;}}_u=\underset{k}{\max }\left\{{\mathrm{\&lambda;}}_{k,\max }\right\};$

Ii. order solve ρ according to formula (12) _k, and calculate total pilot power

If iii. P _all> P, makes λ _l=λ ₀, otherwise make λ _u=λ ₀;

Iv. jump to ii. until meet P-P _all< ∈, wherein ∈ is the limits of error;

In said process, if the pilot power value calculating for user k is less than zero, be set to zero.

2. pilot frequency power optimizing distribution method according to claim 1, is characterized in that, described step 7) in adopt binary search can reduce in the following way mean iterative number of time:

A. by { λ _{k, max}arrange from small to large order ${\mathrm{\&lambda;}}_l=\underset{k}{\max }\left\{{\mathrm{\&lambda;}}_{k,\min }\right\}$ With ${\mathrm{\&lambda;}}_u=\underset{k}{\min }\left\{{\mathrm{\&lambda;}}_{k,\max }\right\};$

B. get successively { λ _{k, max}in one value calculate ρ _k, and calculate total pilot power

If c. P _all< P, carries out described step 7 in current scope) in binary search process, otherwise order

λ _l=λ _u, λ _ufor from { λ _{k, max}extract value, jump to b.

3. pilot frequency power optimizing distribution method according to claim 1 and 2, is characterized in that, described method is the scene for single community.

4. pilot frequency power optimizing distribution method according to claim 1 and 2, is characterized in that, for many cell scenario, first, according to described step 1) to step 5) the formula of the descending achievable rate of process computation:

$\underset{M\&RightArrow;\&infin;}{\lim }{\mathrm{SINR}}_{\mathrm{ik}}=\frac{\frac{E_d}{K}\left(\frac{{\mathrm{\&tau;\&rho;}}_k{\left({\mathrm{\&beta;}}_{\mathrm{ik}}^i\right)}^2}{1+\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_{\mathrm{lk}}^i}\right)}{\underset{j=1,j\&NotEqual;i}{\overset{L}{\mathrm{\&Sigma;}}}\frac{E_d}{K}\left(\frac{{\mathrm{\&tau;\&rho;}}_k{\left({\mathrm{\&beta;}}_{\mathrm{ik}}^j\right)}^2}{1+\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{\mathrm{\&tau;\&rho;}}_k{\mathrm{\&beta;}}_{\mathrm{lk}}^j}\right)+1}$ Formula (13)

Wherein, represent in l community the slow fading coefficient of k user terminal to i base station, think that the user of simultaneously transmitting pilot tone in all communities has identical pilot power, formula (13) is simplified, obtain:

$\underset{M\&RightArrow;\&infin;}{\lim }{\mathrm{SINR}}_{\mathrm{ik}}=\frac{\frac{E_d}{K}{\mathrm{\&rho;}}_k{\left({\mathrm{\&beta;}}_{\mathrm{ik}}^i\right)}^2}{{\mathrm{\&rho;}}_k(\underset{j=1,j\&NotEqual;i}{\overset{L}{\mathrm{\&Sigma;}}}\frac{E_d}{K}{\left({\mathrm{\&beta;}}_{\mathrm{ik}}^j\right)}^2+\underset{l=1}{\overset{L}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}_{\mathrm{lk}}^i)+1}$ Formula (14)

Then, according to described step 6) and step 7) mode construct lagrangian optimization problem, and obtain the optimal value of pilot power by binary search.